Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Claude Haiku 4.5 is clearly ahead on the aggregate, 61 to 57. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Claude Haiku 4.5 is also the more expensive model on tokens at $0.80 input / $4.00 output per 1M tokens, versus $0.30 input / $1.20 output per 1M tokens for MiniMax M2.7. That is roughly 3.3x on output cost alone.
Pick Claude Haiku 4.5 if you want the stronger benchmark profile. MiniMax M2.7 only becomes the better choice if agentic is the priority or you want the cheaper token bill.
Claude Haiku 4.5
51.9
MiniMax M2.7
57
Claude Haiku 4.5
53.1
MiniMax M2.7
56.2
Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.
Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.
Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.
Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.
Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.
Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.
Claude Haiku 4.5 is ahead overall, 61 to 57. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 41% and 57%.
MiniMax M2.7 has the edge for coding in this comparison, averaging 56.2 versus 53.1. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.
MiniMax M2.7 has the edge for agentic tasks in this comparison, averaging 57 versus 51.9. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.